Psychiatrists and psychologists are increasingly tasked by the courts to lend their expertise in the determination of future dangerousness of a defendant. Although sentencing is the most common setting for expert testimony regarding violence risk, mental health professionals may be called upon at all stages of the jurisprudence process. Within the federal court system, determination of dangerousness is requested when a defendant is found not guilty by reason of insanity or incompetent to stand trial and not restorable. Similarly, when a mentally ill person is deemed dangerous to others and is consequently committed to a secure treatment facility, their dangerousness is reevaluated by the committing court at regular intervals (eg, annually). Dangerousness assessments are also common when a sentenced inmate with history of violence reaches the end of their sentence.
Dangerousness is a purely legal construct concerned with the prediction of a binary outcome—whether a person will re-offend—an assumption that does not accommodate current scientific understanding of human behavior. Behavioral science experts operate from an understanding that actions are influenced by a multitude of salient and unknown factors at any given moment and, therefore, base their opinions on relative risk probabilities. Unlike prediction, which implies definitive judgment about future behavior and is incorrect if a behavior does not occur, risk assessment involves the estimation of the likelihood of a behavior based on a combination of factors known to be important in its genesis, and as such need not materialize to be valid. Violence risk assessment approaches fall into three broad categories: unstructured clinical opinion, actuarial assessment, and structured professional judgment. This article critically reviews the theory and methodology behind each approach, introduces the reader to several structured assessment instruments commonly used in forensic practice, and presents research evidence pertaining to the predictive power of various instruments. An illustrative case is used to demonstrate the various approaches in an applied setting.
The defendant is a 28-year-old man charged with aggravated armed robbery. That the defendant brandished a firearm in the commission of the crime served as the aggravating factor, although no bodily harm came to any party. The defendant has prior convictions for possession of cannabis with intent to deliver, mob action, and illegal possession of a firearm. The court is considering pretrial release for the defendant and requests expert opinion to aid in the determination of future dangerousness prior to granting conditional release.
Types of Risk Assessments
Unstructured Clinical Approach
The unstructured clinical approach to risk assessment imposes no constraints or guidelines on the examiner, allowing for an idiographic, or person-centered, analysis of offenders' behavior and a context-specific tailoring of risk management strategies. Proponents of the unstructured approach stress the unique value of clinical judgment in risk assessment, and argue that structured approaches preclude the consideration of important interpersonal cues in favor of broad artificial criteria that may not apply in a specific case. The unstructured approach, however, has been widely criticized for lacking reliability, validity, and accountability.1 Critics argue that any assessment technique that relies exclusively on professional discretion is based more on the preferences and biases of the evaluator than comprehensive consideration of the relevant criminogenic risk factors. Indeed, cognitive psychology literature teaches us that human judgment is vulnerable to a number of biases, such as the fundamental attribution error (tendency to overemphasize stable dispositional traits and devalue situational factors), illusory correlation (belief in a relationship between variables despite contrary empirical evidence), and tendency to assign undue weight to certain factors because of recency (eg, most current arrest) or salience (eg, media coverage of the crime). Cultural differences between the examiner and the examinee have also been shown to affect clinical judgment.2 Cognitive biases are further complicated by the fact that a judgment of non-dangerousness carries greater potential professional consequences for the examiner than a judgment of dangerousness.
Arguably the first legal test of unstructured professional testimony was the 1983 Supreme Court case that challenged the admissibility of expert opinion regarding future dangerousness by a prosecution expert.3 Thomas Barefoot was sentenced to die by a Texas jury for the 1978 murder of a police officer, in part because a psychiatrist, Dr. James Grigson, testified that there was a “one hundred percent and absolute” chance that Barefoot would commit other violent criminal acts in the future. Barefoot appealed to the Supreme Court alleging undue judicial deference to expert testimony. The American Psychiatric Association (APA) filed an amicus curiae brief in support of Barefoot's position, stating “Psychiatric testimony based on hypothetical data is unreliable due to inherent limitations of current psychiatric clinical and experimental knowledge and practice.” The Supreme Court upheld the death sentence despite the APA position, and Barefoot was executed by lethal injection in 1984.3
Structured Risk Assessment
The Barefoot case prompted outrage within the psychiatric community and gave impetus to research on violence risk prediction. In the ensuing 3 decades, studies have identified a number of criminogenic variables that have been standardized into structured instruments designed to allow for empirically based assessment of violence risk. Two types of structured risk assessment schemes have emerged: actuarial prediction and structured professional judgment; both are reviewed below.
Actuarial prediction. Actuarial measures are constructed by prospectively following a sample of offenders in the community and identifying combinations of variables that evidence significant statistical associations with a target outcome (eg, general, violent, or sexual reoffending). The items are weighted according to their predictive power, and the total score can then be used to estimate a statistical probability associated with reoffending in the standardization sample. Several commonly used actuarial instruments are outlined in Table 1. Of these, the Violence Risk Appraisal Guide (VRAG)4 is the most widely accepted. It includes 12 mostly static items (eg, psychopathy checklist score, history of elementary school maladjustment, failure on prior conditional release) derived from a sample of more than 1,200 offenders released from forensic psychiatric facilities in Ontario, Canada. Mean time to re-offense was 81 months, and violent recidivism was defined as a new charge for homicide, attempted homicide, kidnapping, forcible confinement, assault with injury, or rape. Based on a clinical interview, each item is assigned a numerical rating according to the degree of applicability to the examinee, the total score being associated with a 6-year or a 10-year probability of re-offending in the standardization sample.
Actuarial Risk Assessment Instruments in Common Practice
Structured professional judgment. In contrast to actuarial measures, structured professional judgment tools (Table 2) do not involve assigned numerical values or assess statistical probability of reoffending. Items are drawn from existing research and professional literature on violence and recidivism and are combined into checklists designed to help the examiner structure the clinical assessment of risk in accordance with relevant criminogenic variables. The Historical, Clinical, and Risk Management (HCR)-20 (now in its third revision) is arguably the most widely accepted of these instruments, and includes 10 historical, 5 clinical, and 5 risk management variables.5 Many of the historical variables parallel those of the VRAG, including psychopathic personality traits, substance abuse, and prior antisocial behavior. Clinical variables include level of insight, violent attitudes, and major psychiatric illness. Risk management variables include dynamic factors that may serve as targets of intervention such as poor coping skills and inadequate social supports. A supplemental instrument to the HCR-20, the Structured Assessment of Protective Factors for Violence Risk (SAPROF),6 was designed to further guide rehabilitative strategies by identifying dynamic factors associated with criminal desistance in offenders with histories of violence.
Structured Professional Judgment Instruments in Common Practice
Application of Risk Assessment Methods
We return here to the Illustrative Case. At his interview, the defendant presented as an anxious and interpersonally inhibited young man of diminutive stature. He described a troubled upbringing in a large, urban, public housing development, including chronic exposure to violence. He had never met his father. His mother worked two shift jobs, leaving the care of the defendant and his four siblings in the hands of his grandmother. He did well academically in school, but was expelled in 11th grade for drug possession, fighting, and excessive truancy. He joined a street gang at age 14 years because he “wanted to belong.” Part of his initiation was a drive-by shooting of a rival gang member; he “shot in his direction” but was unsure if this resulted in bodily injury. He demonstrated genuine remorse for his actions and reported intrusive rumination about the victims. He expressed a desire to pursue a general equivalency degree and legal employment, but did not know where to begin, stating “the hood is all I've ever known.” The defendant obtained a score of 21 on the Psychopathy Checklist–Revised, which falls below the cutoff suggested for diagnosis of psychopathy. His VRAG-Revised score was 10; in the standardization sample, 66.1% of offenders with this score had violently reoffended within 6 years of release. He endorsed several violence risk factors on the HCR-20 including prior history of antisocial and violent behavior, substance abuse, and inadequate social supports. Conversely, he endorsed several protective factors on the SAPROF, including empathy, motivation for treatment, and prosocial attitudes toward authority. The examiner opined that the defendant had a mild-to-moderate risk of violent reoffending within the next 5 to 10 years, which could be mitigated with targeted interventions including intensive case management, halfway housing in a different neighborhood, substance abuse treatment, and participation in educational and job training programs. At the time of the 2-year reassessment, the defendant was progressing well toward his goals with no new arrests.
As demonstrated by the Illustrative Case, each approach brings unique contributions to violence risk assessment. Still, many clinicians continue to prefer unstructured assessment, believing that the structured approaches (particularly the actuarial method) rob the evaluation process of the rich context of clinical observation. Certain skepticism regarding the ability of standardized methodology to improve insight into human behavior has a longstanding tradition among mental health practitioners. “To reason … in terms of scores and score-patterns is to do violence to the nature of the raw material” said Schafer7 in his seminal opus on the use of psychometric testing in clinical practice. Indeed, although most clinicians regard the position that expert testimony should be limited exclusively to statistical estimation as extreme, actuarial tools are well liked by the courts because they allow the trier of fact to conceptualize risk in terms of tangible numerical probability. To avoid compounding judicial deference to expert opinion with the illusory certainty of statistical correlation, the evaluator should heed, and communicate to the fact finder, the limitations of actuarial assessment. For example, actuarial measures are not strictly informative regarding individual risk but estimate the probability of reoffending within geographically, demographically, and criminologically varied standardization samples, meaning that the validity of actuarial estimation is contingent on the match between the standardization sample and the examinee.
The structured professional judgment approach, by virtue of relying on professional expertise with a structured application, attempts to minimize the limitations of unstructured clinical and actuarial assessment while retaining the strengths of each. Structured professional measures have other advantages, such as applicability across different offender samples and ability to guide intervention. However, unlike actuarial assessment, effective application of structured professional judgment requires familiarity with research findings regarding relevant predictors of a target behavior.
How do the different risk assessment approaches compare in terms of their predictive accuracy? One unpublished meta-analysis of 166 studies with more than 2,600 participants found that prediction of general recidivism based on unstructured clinical judgment was little better than chance (area under the cover [AUC] = 0.58), whereas actuarial and structured professional judgment approaches evidenced adequate classification accuracy with AUC estimates of 0.67 and 0.70, respectively (L. S. Guy, PhD, unpublished data, 2017). Although the predictive validity of unstructured clinical approaches has not been subjected to rigorous systematic testing (the above-mentioned analysis included only six studies in this category), subsequent large scale meta-analyses involving participants across Europe and North America have replicated adequate-to-good classification accuracy for actuarial and structured professional judgment instruments, with AUCs ranging from 0.68 to 0.74 for violent reoffending.8,9 In 2011, the American Psychological Association filed an amicus curiae brief in the US Supreme Court case of Coble v Texas,10 which challenged the admission of unstructured psychiatric testimony regarding future dangerousness in a death penalty case. In an echo of the Barefoot v Estelle opinion,3 the APA position was that unstructured clinical testimony should not be used in dangerousness determination as it is not based on science and is unduly persuasive to juries. Conversely, in considering 3 decades of violence prediction research, the APA affirmed that structured risk assessment methods are accepted by the scientific community as valid and reliable means to inform determinations of future dangerousness in a variety of contexts.
Although structured risk assessment instruments undoubtedly represent an improvement over unstructured clinical opinion, scores on these measures explain only 40% to 55% of inter-person variance in violent reoffending,8,9 and one recent meta-analysis reported substantial variability in actual rates of violent recidivism in offenders classified as high-risk by structured instruments.11 Clearly, much work is yet to be done to improve the classification accuracy of structured violence prediction tools. Several research initiatives are ongoing, including efforts to examine the validity and reliability of structured risk assessment in female offenders, and to identify cognitive, affective, and biological markers for violent behavior that could improve the predictive power of extant instruments.
Ultimately, given the nascent state of scientific understanding of human behavior in general, the ability to predict violent behavior with absolute accuracy is still far in our future. Given the extremely high personal cost associated with inaccurate classification of dangerousness, experts have advocated for the use of multiple approaches, including the assessment of protective factors, as a standard of practice when evaluating violence risk.12 In the age of heuristic thought and pervasive misinformation, the task of the mental health expert within the jurisprudence system is not only to provide independent expert testimony in a specific case, but also to educate the courts on the broad findings from behavioral science. By remaining informed of current research findings and recognizing the strengths and limitations of the various methodologies, the forensic examiner will be well equipped to provide the courts with valid and reliable assessment of violence risk.
- Litwack TR. Actuarial versus clinical assessments of dangerousness. Psychol Public Policy Law. 2001;7(2):409–443. doi:10.1037/1076-8922.214.171.1249 [CrossRef]
- Lezak MD, Howieson DB, Bigler ED, Tranel D. Neuropsychological Assessment. 5th ed. New York, NY: Oxford University Press; 2012.
- Barefoot v Estelle, 103 SCt, 3383 (1983).
- Harris GT, Rice ME, Quinsey VL. Violence Risk Appraisal Guide-Revised, 2013: User Guide. Kingston, Ontario: Data Services, Queen's University Library; 2016.
- Douglas KS, Shaffer C, Blanchard AJE, Guy LS, Reeves K, Weir J. HCR-20 Violence Risk Assessment Scheme: Overview and Annotated Bibliography. Burnaby, Canada: Mental Health, Law, and Policy Institute, Simon Fraser University; 2014.
- de Vogel V, de Vries Robbe M, de Ruiter C, Bouman YHA. Assessing protective factors in forensic psychiatric practice: introducing the SAPROF. Int J Forensic Ment Health. 2011;10(3).171–177. doi: . doi:10.1080/14999013.2011.600230 [CrossRef]
- Schafer R. The Clinical Application of Psychological Tests. New York, NY: International Universities Press; 1948.
- Yang M, Wong SCP, Coid J. The efficacy of violence prediction: a meta-analytic comparison of nine risk assessment tools. Psychol Bull. 2010;136(5):740–767. doi: . doi:10.1037/a0020473 [CrossRef]
- Singh JP, Grann M, Fazel S. A comparative study of violence risk assessment tools: a systematic review and metaregression analysis of 68 studies involving 25,980 participants. Clin Psychol Rev. 2011;31(3):499–513. doi: . doi:10.1016/j.cpr.2010.11.009 [CrossRef]
- Coble v Texas, 131 SCt, 3030 (2011).
- Singh JP, Fazel S, Gueorguieva R, Buchanan A. Rates of violence in patients classified as high risk by structured risk assessment instruments. Br J Psychiatry. 2014;204(3):180–187. doi: . doi:10.1192/bjp.bp.113.131938 [CrossRef]
- Conroy MA, Murrie DC. Forensic Assessment of Violence Risk: A Guide for Risk Assessment and Risk Management. Hoboken, NJ: John Wiley & Sons; 2007. doi:10.1002/9781118269671 [CrossRef]
- Andrews DA, Bonta J. The Level of Service Inventory–Revised. Toronto, ON: Multi-Health Systems; 1995.
- Monahan J, Steadman HJ, Appelbaum PS, et al. The classification of violence risk. Behav Sci Law. 2006;24(6):721–730. doi: . doi:10.1002/bsl.725 [CrossRef]
- Hilton NZ, Harris GT, Rice ME, Lang C, Cormier CA, Lines KJ. A brief actuarial assessment for the prediction of wife assault recidivism: the Ontario domestic assault risk assessment. Psychol Assess. 2004;16(3):267–275. doi: . doi:10.1037/1040-35126.96.36.1997 [CrossRef]
- Quinsey VL, Harris GT, Rice ME, Cormier CA. Violent Offenders: Appraising and Managing Risk. 2nd ed. Washington, DC: American Psychological Association; 2006.
- Hanson RK, Helmus L, Thornton D. Predicting recidivism amongst sexual offenders: a multi-site study of Static-2002. Law Human Behav. 2010;34(3):198–211. doi:10.1007/s10979-009-9180-1 [CrossRef]
- Duwe G, Freske P. The Minnesota Sex Offender Screening Tool-3.1 (MnSOST-3.1). In: Phenix A, Hoberman HM, eds. Sexual Offending: Predisposing Antecedents, Assessments and Management. New York, NY: Springer; 2016:489–502. doi:10.1007/978-1-4939-2416-5_22 [CrossRef]
- Kropp PR, Gibas A. The Spousal Assault Risk Assessment guide (SARA). In: Otto RK, Douglas KS, eds. International Perspectives on Forensic Mental Health. New York, NY: Routledge/Taylor & Francis Group; 2010:227–250.
- Rettenberger M, Boer DP, Eher R. The predictive accuracy of risk factors in the Sexual Violence Risk-20 (SVR-20). Criminal Justice Behav. 2011;38(10):1009–1027. doi: . doi:10.1177/0093854811416908 [CrossRef]
- Hoge RD. Youth level of service/case management inventory. In: Otto RK, Douglas KS, eds. International Perspectives on Forensic Mental Health. New York, NY: Routledge/Taylor & Francis Group; 2010:81–98.
- Borum R, Lodewijks H, Bartel PA, Forth AE. Structured Assessment of Violence Risk in Youth (SAVRY). International Perspectives on Forensic Mental Health. New York, NY: Routledge/Taylor & Francis Group; 2010:63–80.
Actuarial Risk Assessment Instruments in Common Practice
|Level of Service Inventory–Revised13
||Adult male and female inmates and probationers in Canada and US
|Violence Risk Appraisal Guide–Revised4
||General and sexual violent reoffending
||Adult male offenders in Canada
|Classification of Violence Risk14
||Adult male and female psychiatric patients in the US
|Ontario Domestic Assault Risk Assessment15
||Intimate partner violence
||Canadian men arrested for a domestic violence offense
|Sex Offender Risk Appraisal Guide16
||Violent sexual reoffending
||Adult male sex offenders in Canada
||Adult male sex offenders from the US, Canada, New Zealand, and Europe
|Minnesota Sex Offender Screening Tool18
||Adult male sex offenders released from Minnesota prisons
Structured Professional Judgment Instruments in Common Practice
|Historical, Clinical, and Risk Management-20, third version5
||Adult offenders and psychiatric patientsa
|Spousal Assault Risk Assessment19
||Intimate partner violence
||Adult males with history of domestic violence
|Sexual Violence Risk-2020
||Adult male offenders
|Youth Level of Service/Case Management Inventory21
||Male and female adolescent offenders
|Structured Assessment for Violence Risk in Youth22
||Male and female adolescent offenders
|Structured Assessment of Protective Factors for Violence Risk6
||Desistance from general violence
||Adult male and female offenders